(Hmm, it's in November, thought it was earlier)
There is one at 25th of september, Lichtmis, Meppel 😀
grtz
ST
There is one at 25th of september, Lichtmis, Meppel 😀
grtz
ST
Yes, Lichtmis; but that's quite far from where I live and a little bit smaller. But from what I've heard it's good too.
Hm, I have family living in Drachten, might be worthwhile to combine a visit to them with Meppel...
Also there's a nice one every year in Belgium just south of Brussels (in La Louviere).
Also there's a nice one every year in Belgium just south of Brussels (in La Louviere).
I've finally been able to make some simulations.
The process of simulating ESL's is still quite slow, and not very precise. Especially at high frequencies/large panel sizes the simulation is quite slow.
But I can still improve the simulation in many ways to make it faster and much more detailed.
For the people interested: It turned out to be something like a very simple boundary-element-method simulation of the Helmholtz-equation. (after I found how it's called, I've found a lot of ideas to improve the simulation, but I don't know of how much added value it would be, considering the time it would take.
The program is entirely written in C, and the calculation of the animated plot takes around 60 minutes of time on an AMD3500+ CPU. GPU acceleration would improve it a little, but there are a lot of algorithm improvements possible, which make more difference.
What physics did i put into the simulation? The ESL is simulated as a "flat" pressure source, with a diaphragm on it. The diaphragm has a weight and a stiffness-constant. The pressure-source and diaphragm is divided in (21x21=)441 points, for which everything is calculated.
The wave equation itself has not been applied to the diaphragm, only to the air. So only the lowest frequency resonance-mode of the diaphragm is calculated. But you can add them manually by making multiple simulations with different stiffnes-constants of the membrane, and then adding everything together (with some weight-factors). Doing it this way makes the results much more reliable, given the small amount simulated points.
It should be possible to simulate ~10.000 points with beter programming, but still simulating waves on the surface wouldn't be very accurate.
Here are some of the plots. It's all with 441 simulating points on membrane of 20x20cm. The weight is 10g/m^2; the stiffness is 10.000N/m (~spacers at 4cm, with a weight of 1.25kg per meter at the sides)
The whole electrical side of things I didn't yet implement, but that's the easy part of this. So at the moment I can't calculate the impedance / efficiency / maximum pressure etc etc.
These are a series plots of the sound pressure-plots with frequency slowly increasing from 18Hz up to around 20kHz (highest frequencies are a little less accurate). The sound pressure is calculated in front of the ESL up to distances of 4m, here you look 'on top of' the ESL, radiating in free space.
Here a plot at different distances: 0.5,1,2 and 4 meters, at the heartline from the ESL. (in decibels).
Same plot, now at ~20 degrees:
Note that the resonance is VERY strong. I don't think it'll ever be that large in reality. I could easily put some damping in the model, but I don't know what values would be realistic, so I've left it pure.
If someone is interested in source-code, feel free to PM me.
The process of simulating ESL's is still quite slow, and not very precise. Especially at high frequencies/large panel sizes the simulation is quite slow.
But I can still improve the simulation in many ways to make it faster and much more detailed.
For the people interested: It turned out to be something like a very simple boundary-element-method simulation of the Helmholtz-equation. (after I found how it's called, I've found a lot of ideas to improve the simulation, but I don't know of how much added value it would be, considering the time it would take.
The program is entirely written in C, and the calculation of the animated plot takes around 60 minutes of time on an AMD3500+ CPU. GPU acceleration would improve it a little, but there are a lot of algorithm improvements possible, which make more difference.
What physics did i put into the simulation? The ESL is simulated as a "flat" pressure source, with a diaphragm on it. The diaphragm has a weight and a stiffness-constant. The pressure-source and diaphragm is divided in (21x21=)441 points, for which everything is calculated.
The wave equation itself has not been applied to the diaphragm, only to the air. So only the lowest frequency resonance-mode of the diaphragm is calculated. But you can add them manually by making multiple simulations with different stiffnes-constants of the membrane, and then adding everything together (with some weight-factors). Doing it this way makes the results much more reliable, given the small amount simulated points.
It should be possible to simulate ~10.000 points with beter programming, but still simulating waves on the surface wouldn't be very accurate.
Here are some of the plots. It's all with 441 simulating points on membrane of 20x20cm. The weight is 10g/m^2; the stiffness is 10.000N/m (~spacers at 4cm, with a weight of 1.25kg per meter at the sides)
The whole electrical side of things I didn't yet implement, but that's the easy part of this. So at the moment I can't calculate the impedance / efficiency / maximum pressure etc etc.
These are a series plots of the sound pressure-plots with frequency slowly increasing from 18Hz up to around 20kHz (highest frequencies are a little less accurate). The sound pressure is calculated in front of the ESL up to distances of 4m, here you look 'on top of' the ESL, radiating in free space.
An externally hosted image should be here but it was not working when we last tested it.
Here a plot at different distances: 0.5,1,2 and 4 meters, at the heartline from the ESL. (in decibels).
An externally hosted image should be here but it was not working when we last tested it.
Same plot, now at ~20 degrees:
Note that the resonance is VERY strong. I don't think it'll ever be that large in reality. I could easily put some damping in the model, but I don't know what values would be realistic, so I've left it pure.
An externally hosted image should be here but it was not working when we last tested it.
If someone is interested in source-code, feel free to PM me.
Dear Calvin,
can you elaborate more on 0.2mm tolerances. One could achieve such a figure for let's say a d/s spacing.
My experience says, that even with the structure similar to prestressed concrete, 100 cm long piece of a reasonable cross section,
even from annealed and aged metal right out of the milling machine will bend/warp under the residual stress/own weight to a figure higher than that.
Use of stiffer material like high carbon steel/duralumin complicates the matter even more.
Honeycomb structure of Quads, for instance, will never qualify to such accuracy.
- Status
- Not open for further replies.